1. Field of the Invention
This invention relates to a modulating device, demodulating device and transmission medium, and in particular to a modulating device, demodulating device and transmission medium which are suitable for modulating data for data transmission or recording on a recording medium, and demodulating the modulation code obtained by this modulation so as to reproduce the data.
2. Description of the Related Art
When data is transmitted on a predetermined transmission path or recorded for example on recording media such as magnetic disks, optical disks, and magneto-optical disks, data modulation is performed which is suitable for transmission or recording.
One such type of modulation is known as block coding This block coding converts a data sequence to blocks of m.times.i bit units (referred to hereafter as data words), and this data word is converted to a code word comprising n.times.i bits according to a suitable code rule
When i=1, this code is a fixed length code expressed by (d,k;m,n;1). When plural i are selected, a predetermined i is selected from the range 1 to imax (maximum value of i) and the conversion is performed, the code is a variable-length code. This block encoded code is represented by a variable length code (d,k;m,n;r).
Here, i is known as a restriction length, and imax is r (the maximum restriction length) . The minimum run d shows the minimum number of consecutive "0"s in repeated "1"s in a code sequence. The maximum run k shows the maximum number of consecutive "0"s in repeated "1"s in a code sequence.
In compact disks or mini-discs (trademark) etc., NRZI (Non Return to Zero Inverted) modulation, wherein "1" means inversion and "0" means non-inversion, is performed on the variable length code obtained as above, and the NRZI modulated variable length code (hereafter, referred to as a recorded waveform sequence) is recorded.
Various modulation techniques have been proposed. If the minimum inversion interval of the recorded waveform sequence is Tmin(=d+1), and the maximum inversion interval is Tmax(=k+1), to record at a high density in a linear velocity direction, the minimum inversion interval Tmin should be long, that is, the minimum run d should be large. From the clock reproduction aspect, moreover, the maximum inversion interval Tmax should be short, that is, the maximum run k should be small.
One modulation technique used by magnetic disks or magneto-optical disks, etc., is RLL(1-7). The parameters of this modulating technique are (1,7;2,3;2). The minimum inversion interval Tmin is 2(=1+1) T (=(2/3).times.2Tdata=1.33Tdata). The maximum inversion interval Tmax is 8(=7+1) T(=(2/3)X8Tdata=5.33Tdata). In addition, a detection window width Tw(=(m/n)XT) is 0.67(=2/3) Tdata.
For example, the variable length conversion table for the RLL(1-7) code is as follows:
TABLE 1 ______________________________________ RLL (1,7;2,3;2) data code ______________________________________ i=1 11 00x 10 010 01 10x i=2 0011 000 00x 0010 000 010 0001 100 00x 0000 100 010 ______________________________________
The symbol x in the conversion table is 1 when the next channel bit is 0, and 0 when the next channel bit is 1. The restriction length r is 2.
However, the RLL(1-7) code can be formed from fixed length codes. It may be expected that modulation/demodulation will be easier if fixed length codes are used. For example, as the output is always a fixed number of bits, i.e. data words are output two bits at a time, and code words are output three bits at a time, the construction is simple.
The fixed length conversion table for the RLL(1-7) code is as follows. Herein, to distinguish it from the aforesaid variable-length, the fixed length table will be represented by RLL-F(1-7).
The parameters of RLL-F(1-7) are (1,7;2,3;1), and if the recorded waveform sequence bit interval is T, the minimum inversion interval Tmin is 2 (=1+1) T. If the data sequence bit interval is Tdata, the minimum inversion interval Tmin is 1.33(=(2/3).times.2) Tdata. Moreover, the maximum inversion interval Tmax is 8T (5.33Tdata). Further, the detection window width Tw is given by (m/n) xT, and its value is 0.67(=2/3) T. The RLL-F(1-7) table is shown in Table 2. This table is an ISO standard table.
TABLE 2 ______________________________________ RLL-F (1,7;2,3;1) Immediately preceding Current Next Converted code word data word data word code word ______________________________________ x 00 0x 001 0 00 1x 000 1 00 1x 010 0 01 0x 001 0 01 1x 000 1 01 00 010 1 01 not 00 000 0 10 0x 101 0 10 1x 010 0 11 00 010 0 11 not 00 100 ______________________________________
"x" in the above conversion table shows that either 0 or 1 is acceptable. not00 means any of the data words 01, 10 and 11.
This RLL-F(1-7) (1,7;2,3;l) may also be obtained by replacing converted data words as in Table 3. Further, demodulation can be performed 1:1 as in the above Table 2.
TABLE 3 ______________________________________ RLL-F (1,7;2,3;1) (2) Immediately preceding Current Next Converted code word data word data word code word ______________________________________ 0 00 11 010 0 00 not 11 100 0 01 0x 010 0 01 1x 101 0 10 0x 000 0 10 1x 001 1 10 11 010 1 10 not 11 000 0 11 0x 000 1 11 0x 010 x 11 1x 001 ______________________________________
`x` in the aforesaid conversion table shows that either 0 or 1 is acceptable. Moreover, not11 means any of the data words 00, 01 and 10. However, considering a T distribution of a channel bit sequence which is modulated by the aforesaid (1-7) codes, the occurrence frequency of 2T which is Tmin is the greatest, 3T, 4T, and 5T are less, and the occurrence frequency of 8T is the least. In general, if a large amount of edge information occurs early as in the case of 2T and 3T, this is advantageous for clock reproduction. However, when on the other hand 2T occurs repeatedly, the output waveform amplitude is smaller than for large T such as for example 5T and 6T. This is because the output during playback is smaller the higher the region depending on the optical characteristics of the lens.
In the RLL(1-7) code, 2T is output the most toward the high region. When a minimum mark is repeatedly recorded at a high linear density, its reproduction output is small, so patterns with poor S/N increase, and this makes signal detection unstable.
Further, for example, when high region playback characteristics deteriorate due to, for example, defocusing or inclination in a tangential direction, clock reproduction is expected to become more unstable.
RLL(1-7) is often combined with PRML (Partial Response Maximum Likelihood), to improve S/N during playback of a high density recording. This method comprises, for example, Viterbi decoding equalized by PR (1,1) or PR(1,2,1) by matching the RF reproduction waveform to media characteristics. For example, a desirable reproduction output when equalization is performed by PR(1,1) is as follows.
______________________________________ 1 0 1 0 0 1 0 (channel bit data swquence) 1 1 0 0 1 0 0 (after NRZI conversion) ....... 1 1 1 1 -1 -1 -1 -1 1 1 1 1 -1 -1 ....... ...+2 0 -2 0 +2 0 ... (reproduction output). ______________________________________
Therefore, when 2T which is Tmin occurs repeatedly, waveform equalization is performed to approach this reproduction output. In general, the waveform interference becomes longer the higher the linear density, therefore, waveform equalization also becomes longer as in PR(1,2,2,1) or PR(1,1,1,1).
However when the minimum run d=1 and a suitable waveform equalization is PR (1,1,1,1) as a result of high linear density, considering a situation when 2T which is Tmin occurs repeatedly, the playback signal at that time is
______________________________________ 1 0 1 0 1 0 1 0 1 0 (channel bit data sequence) 1 1 0 0 1 1 0 0 1 1 (after NRZI conversion) ....... 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 ....... ...0 0 0 0 ... (reproduction output) ______________________________________
and the zero level will be traced for a long time.
This shows that a situation when practically no signal level is output after waveform equalization continues, and causes considerable loss of data reproduction or clock reproduction stability. In addition, merge does not occur while 2T continues even when Viterbi decoding is performed.
This kind of channel bit data sequence where Tmin is repeated, for example in the case of RLL (1,7;2,3;2) in Table 1, occurs when the premodulated data sequence is "10-01-10-01-10- . . . "
Similarly, in the case of RLL-F(1,7;2,3;1) in Table 2 , this occurs when the premodulated data sequence is "10-00-10-00-10- . . ."
In the case of RLL-F(1,7;2,3;1) (2) in Table 3 , this occurs when the premodulated data sequence is "01-11-01-11-01- . . . "
As described hereabove, when recording media such as magnetic disks, magneto-optic disks or optical disks are high density, and when a code such as RLL(1-7) or RLL-F(1-7) is selected as a modulation code, if the minimum inversion interval Tmin occurs too many times in succession, patterns with poor S/N occur repeatedly, so signal detection is unstable which is disadvantageous for clock reproduction.
Similarly, in the case of high linear density, if the minimum inversion interval Tmin occurs repeatedly when PR(1,1,1,1) equalization is performed for the d=1 code, the reproduced signal will be zero for a long time and Viterbi decoding will not merge, which is disadvantageous for clock reproduction.